Studies on automatic recognition of preposition BA´s usages based on statistics

Researching on automatic recognition of preposition´s usages is one of the important contents of the Chinese function words Knowledge Corpus. Aiming at the low recognition probability problem for some high-frequency usages with rule-based method of preposition´ usages recognition, this paper proposed automatically recognizing of preposition´ usages based on statistics. Three statistical models, that is CRF, ME and SVM, were used to label preposition BA´s usages on the tagged corpus of People´s Daily (February, March, April of 2000). The experiment of statistical methods was done for preposition BA. And the final results showed that in general automatic recognition of preposition BA´s usages based on statistics was better than that of rule-based method of preposition BA´s usages recognition.